Upgrade helm-operator to Helm v3

[joelanford]

Description of the change:
This PR bumps Helm to v3.0.0, adds the helm-2to3 plugin to auto migrate existing CRs, and refactors relevant portions of the helm operator code due to the restructuring between helm 2 and helm 3

Motivation for the change:
To support Helm 3 in Operator SDK

[njhale]

We've been wanting to, but some openshift automation added a v3.11 tag to our repository, which would force us to start at a greater semver for minimal version selection to skip it. IIRC, this happened right as we started to use modules (go 1.11 timeframe). We should revisit this soon.

[joelanford]

Oh no! That's a bummer. I don't envy you. However, since there's no go.mod file on the commit tagged with v3.11.0, you might not be totally hosed. See https://github.com/golang/go/wiki/Modules#semantic-import-versioning, specifically item 2 in the "Three Transitional Exceptions" section.

It might be worth trying to add a valid go module semver tag in a throwaway fork at v0.12.0 to see whether you can import at that version without getting bumped "up" to v3.11.0. I'm pretty sure that will work.

Also since v3.11.0 is not a go module (since it is lacking a go.mod file), you may even be able to delete the v3.11.0 tag with little consequences. Don't quote me on that, but that's something else that might be worth trying in a throwaway fork.

[openshift-ci-robot]

@joelanford: The following test failed, say /retest to rerun them all:

Test name	Commit	Details	Rerun command
ci/prow/marker	67e0b80	link	/test marker

Full PR test history. Your PR dashboard. Please help us cut down on flakes by linking to an open issue when you hit one in your PR.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. I understand the commands that are listed here.

[camilamacedo86]

Why is required have releaseHistoryV2 and releaseHistoryV3? Will we keep the compatibility with old versions?
Should it still work with v2 helms?

[joelanford]

This is necessary to handle automatically migrating existing Helm v2 releases to Helm v3. The idea here is that a user with existing CRs and an existing Helm v2-based operator can upgrade to the new Helm 3-based operator and the new operator will handle the transition seamlessly.

[camilamacedo86]

It mainly shows fine. I added a few questions/comments.
Also, it is missing the CHANGELOG and update all link docs to address the users check the refs in the helm3 docs instead of 2.

[camilamacedo86]

Should not it be removed? Why keep this code commented?

[jmccormick2001]

it could be my environment, but I built sdk for this PR, ran the user guide steps and am
seeing the following:

the CR shows size of 2....

replicaCount: 2
resources: {}
securityContext: {}
service:
  port: 8080
  type: ClusterIP
serviceAccount:
  create: true
  name: null
tolerations: []

kind: List
metadata:
resourceVersion: ""
selfLink: ""

but only a single pod and deployment is running....
jeffmc@doppio114 ~/projects/helmtest/nginx-operator> kubectl get pods
NAME READY STATUS RESTARTS AGE
nginx-operator-6c7c94494b-9nc6m 1/1 Running 0 90s
jeffmc@doppio114 ~/projects/helmtest/nginx-operator> kubectl get deploy
NAME READY UP-TO-DATE AVAILABLE AGE
nginx-operator 1/1 1 1 119s

I'm seeing this in the operator log:
{"level":"error","ts":1575416482.6321547,"logger":"controller-runtime.controller","msg":"Reconciler error","controller":"nginx-controller","request":"default/example-nginx","error":"failed to load chart dir: apiVersion 'v2' is not valid. The value must be "v1"","stacktrace":"github.com/go-logr/zapr.

here is my test env:
jeffmc@doppio114 ~/projects/nginx-operator> kubectl version
Client Version: version.Info{Major:"1", Minor:"16", GitVersion:"v1.16.3", GitCommit:"b3cbbae08ec52a7fc73d334838e18d17e8512749", GitTreeState:"clean", BuildDate:"2019-11-13T11:23:11Z", GoVersion:"go1.12.12", Compiler:"gc", Platform:"linux/amd64"}
Server Version: version.Info{Major:"1", Minor:"16", GitVersion:"v1.16.3", GitCommit:"b3cbbae08ec52a7fc73d334838e18d17e8512749", GitTreeState:"clean", BuildDate:"2019-11-13T11:13:49Z", GoVersion:"go1.12.12", Compiler:"gc", Platform:"linux/amd64"}
jeffmc@doppio114 ~/projects/nginx-operator> go version
go version go1.13.1 linux/amd64

[estroz]

LGTM overall, just a bunch of nits and questions.

[jmccormick2001]

seems to test out ok on OCP 4.2, I deployed using the master branch (helm v2), then deleted the nginx-operator, then built a new nginx-operator against this PR code, redeployed the nginx-operator, then updated the CR to add a new nginx pod, seems to work.

[estroz]

/lgtm

[openshift-ci-robot]

New changes are detected. LGTM label has been removed.